week ending 19 JULY 2013

PHYSICAL REVIEW LETTERS

Distribution of Scattering Matrix Elements in Quantum Chaotic Scattering S. Kumar,1,* A. Nock,1 H.-J. Sommers,1 T. Guhr,1 B. Dietz,2 M. Miski-Oglu,2 A. Richter,2 and F. Scha¨fer3 1

Fakulta¨t fu¨r Physik, Universita¨t Duisburg-Essen, Lotharstraße 1, D-47048 Duisburg, Germany 2 Institut fu¨r Kernphysik, Technische Universita¨t Darmstadt, D-64289 Darmstadt, Germany 3 LENS, University of Florence, I-50019 Sesto-Fiorentino, Italy (Received 9 January 2013; published 16 July 2013)

Scattering is an important phenomenon which is observed in systems ranging from the micro- to macroscale. In the context of nuclear reaction theory, the Heidelberg approach was proposed and later demonstrated to be applicable to many chaotic scattering systems. To model the universal properties, stochasticity is introduced to the scattering matrix on the level of the Hamiltonian by using random matrices. A long-standing problem was the computation of the distribution of the off-diagonal scatteringmatrix elements. We report here an exact solution to this problem and present analytical results for systems with preserved and with violated time-reversal invariance. Our derivation is based on a new variant of the supersymmetry method. We also validate our results with scattering data obtained from experiments with microwave billiards. DOI: 10.1103/PhysRevLett.111.030403

PACS numbers: 03.65.Nk, 05.45.Mt, 11.55.m, 24.30.v

A large part of our knowledge about quantum systems comes from scattering experiments. Even in classical wave systems, observables can often be traced back to a scattering process [1]. Important examples stem from nuclear, atomic and molecular physics, mesoscopic ballistic devices, and even from classical wave systems as, e.g., microwave and elastomechanical billiards, as well as from wireless communication [1–31]. Accordingly, the investigation of scattering phenomena has been a subject of major interest from both theoretical and experimental points of view. Here, we focus on the universal features of chaotic scattering systems. The quantity of interest is the scattering matrix (S matrix) which relates the asymptotic initial and final states, and owing to the flux conservation requirement is unitary. The S-matrix elements are given in terms of the Hamiltonian H describing the scattering center [6,16] by Sab ðEÞ ¼ ab i2Way GðEÞWb ;

(1)

where the inverse of the resolvent GðEÞ reads G1 ðEÞ ¼ E1N H þ i

M X

Wc Wcy :

(2)

c¼1

The coupling vectors Wc account for the interaction between the internal states of H and the M open channel states labeled c ¼ 1; . . . ; M where the full system resides asymptotically before, respectively, after the scattering event. This ansatz yields the most general description of any scattering process in which an interaction zone and scattering channels can be identified. Without loss of generality, we may restrict ourselves to a diagonal average S matrix [32], that is, to orthogonal coupling vectors, viz., Wcy Wd ¼ ðc =Þcd ; c, d ¼ 1; . . . ; M [9,16]. 0031-9007=13=111(3)=030403(6)

The chaoticity of the dynamics in the scattering center is taken into account by assuming that H is a random matrix. This is referred to as the Heidelberg approach [6], as distinguished from the Mexico approach [7,8], where stochasticity is introduced in the S matrix itself. In view of the universality conjecture [33], the Hamiltonian H is modeled by the Gaussian ensemble of N N random matrices with the distribution [5,34,35], P ðHÞ / expð ðN=4v2 ÞtrH 2 Þ. Here, v2 fixes the energy scale and the index characterizes the symmetry class, i.e., the invariance properties of the Hamiltonian. We focus on the cases ¼ 1 (Gaussian orthogonal ensemble, GOE) and ¼ 2 (Gaussian unitary ensemble, GUE) which apply to systems with, respectively, preserved and violated time-reversal (T ) invariance [5,34,35], the former being rotationally invariant as well. The universal spectral properties of closed chaotic systems deduced from these ensembles are by now well understood [5]. Those of the effective Hamiltonian associated with a chaotic scattering process, which includes H and the coupling to the exterior have been derived, e.g., in [16]. Its complex eigenvalues can be extracted from the measured resonance spectra [4,36] in general, only in the regimes of isolated and weakly overlapping resonances. Accordingly, a description of the universal properties of the S matrix itself is indispensable. In their pioneering work [9], using the supersymmetry method [37], Verbaarschot et al. calculated the energy correlation function of two S-matrix elements. Further progress in this direction was made in [10] where three- and fourpoint S-matrix correlation functions were evaluated. While these provide rich information about the scattering process, a full statistical description requires the determination of the distributions of the S-matrix elements. Their knowledge yields information about all their moments and is highly desirable also from an experimental viewpoint [4]. The

030403-1

Ó 2013 American Physical Society

PRL 111, 030403 (2013)

PHYSICAL REVIEW LETTERS

distributions and higher moments have been known hitherto only in the limit of a large number M of open channels and a vanishing average S matrix, i.e., in the Ericson regime [4,11] or in a high-loss environment [29,30]. There the real and the imaginary parts of the S-matrix elements are Gaussian distributed. Otherwise the deviations from this behavior are significant due to the unitarity of the S matrix [4,10,38,39]. The complexity involved in the calculations of the correlation functions [9,10,40] indicates that those of the distributions of the S-matrix elements constitute a challenging task. However, this was partially accomplished in [17] where the distribution of the diagonal S-matrix elements was derived. Moreover, in [18], the statistics of transmitted power, viz. jGnm ðEÞj2 , n m, was calculated. These results have been verified in microwave experiments [19–26]. In the present Letter, we provide analytical results for the distributions of the off-diagonal S-matrix elements which could not be computed with the well-established methods [9,10,17]. The novelty of our approach lies in that a nonlinear sigma model is constructed based on the characteristic function associated with the distributions which is the generating function for the moments. In contrast, the standard supersymmetry approach starts from the generating function for the S-matrix correlations. We introduce the notation }s ðSab Þ, with s ¼ 1, 2 to refer to the real and imaginary parts of Sab , respectively. Thus, Eq. (1) yields for the off-diagonal (a b) elements }s ðSab Þ ¼ ððiÞs Way GWb þ is Wby Gy Wa Þ:

(3)

Determining distributions for }s ðSab Þ, which we denote by Ps ðxs Þ, involves the nontrivial task of performing an ensemble average, Z Ps ðxs Þ ¼ d½HP ðHÞðxs }s ðSab ÞÞ; s ¼ 1; 2: (4) We instead first compute the corresponding characteristic function, Z Rs ðkÞ ¼ d½HP ðHÞ expð ik}s ðSab ÞÞ; (5) and then obtain Ps ðxs Þ as the Fourier transform of Rs ðkÞ, 1 Z1 Ps ðxs Þ ¼ dkRs ðkÞ expðikxs Þ: (6) 2 1 Defining the 2N-component vector W and the 2N 2N matrix As as " # " # 0 ðiÞs G Wa ; As ¼ s y ; (7) W¼ Wb 0 iG we rewrite Rs ðkÞ as Z Rs ðkÞ ¼ d½HP ðHÞ expðikW y As WÞ:

(8)

In this form, the ensemble average can not be performed, because As contains the inverse of H. To carry it out, we

week ending 19 JULY 2013

map the statistical model to superspace. We introduce the 2N vectors zT ¼ ½zTa ; zTb and T ¼ ½aT ; bT consisting of complex commuting and anticommuting (Grassmann) variables, respectively. The supervector is constructed in the usual manner [41] as T ¼ ½zT ; T . Using these vectors, and multivariate Gaussian-integral results, we recast the characteristic function as Z i Rs ðkÞ ¼ d½ exp ðW y þ y WÞ 2 Z i : (9) y A1 d½HP ðHÞ exp s 4k 1 y y Here A1 s ¼ 12 As , and W ¼ ½W ; 0 is a 4N vector. The ensemble average can now be done, leading to an enormous reduction in the degrees of freedom. To facilitate it we block diagonalize A1 s by the transformations

2 z ! T þ z; zy ! zy ; ! T ; y ! y ; " # s 0 ðiÞ 1N : T ¼ is 1N 0

(10)

Here, we have used the fact that the complex quantities zðÞ and zy ð y Þ are independent of each other. The Jacobian of the transformation is ð1ÞN 22ð2ÞN . After the application of Eq. (10) we have to distinguish between the two cases. For ¼ 2 we obtain Z i Rs ðkÞ ¼ ð1ÞN d½ exp ðUys þ y WÞ 2 Z i y A 1 ; (11) d½HP ðHÞ exp 4k with the 4N vector Uys ¼ ½is Wby ; ðiÞs Way ; 0; 0 and matrix A 1 ¼ diag½ðG1 Þy ; G1 ; ðG1 Þy ; G1 , T and ¼ ½zT ; T as above. For ¼ 1, we decompose the 2N vector z into its real and imaginary parts x and y (not to be confused with x1 and x2 ) to construct a 4N vector. In addition, we symmetrize the vector using a , b along with a , b , thereby doubling its size as well. The associated Jacobian equals 22N and, thus, cancels that of the transformation Eq. (10). The 8N supervector is given as y ¼ ½xTa ; yTa ; xTb ; yTb ; ay ; aT ; by ; bT and Z Rs ðkÞ ¼ ð1ÞN d½ expðiy Vs Þ Z i y A 1 : (12) d½HP ðHÞ exp 4k sþ1 þ In this case, VTs ¼ ½is w ws ; wþ s ; i s ; iws ; 0; 0; 0; 0, s T T where ws ¼ ððiÞ Wa Wb Þ=2, and A 1 ¼ 1 y 1 1 y 1 diag½ðG Þ ; G ; ðG Þ ; G 12 . Since the matrix A 1 is block diagonal in both Eqs. (11) and (12), the ensemble averaging is now straightforward.

030403-2

PRL 111, 030403 (2013)

PHYSICAL REVIEW LETTERS

Next, we use the Hubbard-Stratonovitch identity [37,41] to map the integral over the 8N= supervector to a matrix integral in superspace involving an 8=-dimensional supermatrix of appropriate symmetry. This yields Z i 2 Rs ðkÞ ¼ d½ exp r str str ln Fs ; 2 4 M X i ¼ E 1N þ L Wc Wcy ; (13) 4k c¼1

E 1 ; E ¼ 4k 8= with str denoting the supertrace. Here r ¼ ð42 k2 NÞ=v2 , and L ¼ diagð1; 1; 1; 1Þ 12= . Fs equals VTs L1=2 1 L1=2 Vs for ¼ 1, and y 1=2 1 1=2 L W for ¼ 2 with L ¼ L 1N . The Us L supersymmetric representation, Eq. (13), constitutes one of our key results. The orthogonality of Wc leads to M X ic 1 L ; str ln 18= þ str ln ¼ N str lnE þ 4k E c¼1 1 1 ¼ 1 E 1N E

þ

M X c¼1

ðcÞ

M X Wc Wcy c c¼1

W Wcy ; c c

(14)

with ðcÞ ¼ ðE þ ic =ð4kÞLÞ1 . Furthermore, Fs equals a linear combination of matrix elements of ðcÞ multiplied with c , where c ¼ a, b. In Eq. (14), the first term is of order N while the rest is of order 1. Thus, in order to perform the limit N ! 1, we may apply the saddle point approximation. This leads to a separation of into Goldstone modes G and massive modes [41]. The integrals over the latter, being Gaussian ones, can be readily done and yield unity. Therefore, we are left with an expression involving only the Goldstone modes, and consequently, our sigma model reads Z i Rs ðkÞ ¼ dðG Þ exp Fs 4 M Y ic 1 ð=2Þ L ; sdet (15) 18= þ 4k E c¼1 with sdet denoting the superdeterminant and replaced by G in all the ingredients of Eq. (13). In order to perform the remaining integrals, we proceed as in [9,41] and express G in terms of an 8=-dimensional supermatrix Q as G ¼ ðE=8kÞ18= ð=8kÞQ with Q2 ¼ 18= . Here, ¼ ð4v2 E2 Þ1=2 with =ð2v2 Þ identified as the celebrated Wigner semicircle. We use the parametrization of Q as in [9,16,42]. For ¼ 2, it involves pseudoeigenvalues 1 2 ð1; 1Þ, 2 2 ð1; 1Þ, angles 1 , 2 2 ð0; 2Þ, and four Grassmann variables. For ¼ 1, we

week ending 19 JULY 2013

have three pseudoeigenvalues 0 2 ð1; 1Þ, 1 , 2 2 ð1; 1Þ, two O(2) angles 1 , 2 2 ð0; 2Þ, three SU(2) variables m, r, s 2 ð1; 1Þ, and eight Grassmann variables. The product over the superdeterminants in Eq. (15) involves the pseudoeigenvalues only, viz., FO ¼ FU ¼

M Y

gþ c þ 0 þ 1=2 ðgþ þ Þ1=2 2 c c¼1 ðgc þ 1 Þ M Y gþ c þ 2 þ c¼1 gc

þ 1

for ¼ 1; for ¼ 2:

2 2 þ Here g c ¼ ðv c Þ=ðc Þ. gc is related to the transmission coefficient Tc ¼ 1 jScc j2 as gþ c ¼ 2=Tc 1 [9]. The exponential part in Eq. (15) also involves other variables and is quite complicated for ¼ 1. For ¼ 2, the integrals over the Grassmann variables and the angles can be performed and we obtain the same distribution for the real and imaginary parts, Z1 Z1 k2 Rs ðkÞ ¼ 1 d 1 d 2 4ð 1 2 Þ2 1 1 qﬃﬃﬃﬃﬃﬃﬃﬃ F U ð 1 ; 2 Þðt1a t1b þ t2a t2b ÞJ0 k t1a t1b ; (16)

where Jn ðzÞ represents the nth order Bessel function of the qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ first kind, and tjc ¼ j 2j 1j=ðgþ c þ j Þ, j ¼ 1, 2. The ‘‘1’’ in Eq. (16) is an Efetov-Wegner contribution [37] which is essential for the correct normalization, Rs ð0Þ ¼ 1. The distribution function is obtained using Eq. (6) as Ps ðxs Þ ¼ @2 fðxs Þ[email protected] ; Z1 Z1 F U ð 1 ; 2 Þ fðxÞ ¼ xðxÞ þ d 1 d 2 4ð 1 2 Þ2 1 1 ðt1a t1b þ t2a t2b Þðt1a t1b x2 Þ1=2 ðt1a t1b x2 Þ:

(17)

Here, ðuÞ is the Heaviside function. The distributions being identical for s ¼ 1, 2 in this case, the phases have a uniform distribution and the joint density of the real and qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ the imaginary parts depends on x21 þ x22 only. This facilitates the calculation of the distribution of their moduli [43] and those of the cross sections which are given by the squared moduli [4]. This is of particular relevance for the experiments where only these are accessible. For ¼ 1, the calculation involved is rather cumbersome. Nevertheless, we managed to perform all but four integrals. We have Z1 Z1 1 Z1 R1 ðkÞ ¼ 1 þ d 0 d 1 d 2 8 1 1 1 4 Z 2 X d c J ð 0 ; 1 ; 2 ÞF O ð 0 ; 1 ; 2 Þ n kn ; 0

n¼1

(18) where

030403-3

4 ¼ 41 J0 ðk!Þ þ 42 J2 ðk!Þ þ 43 J4 ðk!Þ; with the entries ij given in the Supplemental Material [44]. R2 ðkÞ is obtained by multiplying i to the right-hand side of the expressions for q c in Eq. (19), and changing rc accordingly. The distribution is obtained as @f1 @2 f2 @3 f3 @4 f4 þ 2 þ 3 þ 4 ; @xs @xs @xs @xs f2 ¼ h21 þ 22 ð1 2x2s =!2 Þi;

f3 ¼ h½31 þ 32 ð3 4x2s =!2 Þxs =!i; f4 ¼ h41 þ 42 ð1 2x2s =!2 Þ (20)

Here the angular brackets represent the following:

1 0.8 0.4 0.2 0

-5

0

5

10

-1

15

1

0.5

xs

FIG. 1 (color online). Comparison of analytical and simulation results for ¼ 2: (a) characteristic functions (b) distributions of the real and the imaginary parts of Sab for the choice of parameters M ¼3, E¼0:25, v¼1, 1 ¼ 0:8, 2 ¼1:0, 3 ¼ 1:2, a ¼ 1, b ¼ 2.

obtained with an ensemble of 50 000 random matrices H of dimensions 200 200 from the GUE [5,35]. The agreement is excellent. Unfortunately, there were no experimental data available for this case because a complete T invariance violation could not be achieved. For ¼ 1, we found that Rs ðkÞ is best evaluated using the Efetov variables 0 , 1 , 2 (0 < 0 < , 0 < 1;2 < 1) [37]. These are related to the ’s as 0 ¼ cos 0 and 1;2 ¼ coshð 1 2 Þ. The numerical evaluation of the fourth derivative needed for the computation of Ps ðxs Þ is not feasible. Instead, we determined them with the help of Eq. (6), considering a cutoff for k. This approach works well for a sufficiently flat distribution, whereas, if it is highly localized, it is advantageous to consider the corresponding characteristic function instead. We found that the analytical results converge to the expected Gaussian distributions in the Ericson regime for both values [43]. We tested our analytical results for ¼ 1 with experimental data. To realize a chaotic scattering system, a microwave billiard with the shape of a classically chaotic

1

0

(a)

Analytical Simulation Experimental

R1(k)

-30

(b)

Analytical Simulation Experimental

0.4

0.7

0.2

0.6

0

0.5

Different results for the real and imaginary parts explain their unequal deviations from a Gaussian behavior which was observed in [38,39]. Details of the supersymmetry calculations and further results will be given elsewhere [43]. We evaluated Eqs. (16) and (17) numerically using MATHEMATICA [45]. The corresponding MATHEMATICA codes are included in the Supplemental Material [44]. In Fig. 1, for ¼ 2, we compare the analytical results for characteristic functions and distributions with simulations

0

-0.5

k

Z1 Z1 1 Z1 d 0 d 1 d 2 2 16 1 1 1 Z 2 d c J ð 0 ; 1 ; 2 ÞF O ð 0 ; 1 ; 2 Þ 2hð!2 x2s Þ1=2 ð!2 x2s Þ:

0.6

Ps(k)

0.6 -10

0.9

hhi ¼

Analytical Simulation, P1 Simulation, P2

0.8

þ 43 ð1 8x2s =!2 þ 8x4s =!4 Þi:

-15

R2(k)

f1 ¼ h11 xs =!i;

(b)

1

2 ¼ 21 J0 ðk!Þ þ 22 J2 ðk!Þ;

3 ¼ 31 J1 ðk!Þ þ 32 J3 ðk!Þ;

Ps ðxs Þ ¼ ðxs Þ þ

Analytical Simulation, R1 Simulation, R2

0.6

and of the complex conjugate of q c , rc ¼ ðqc Þ , and pﬃﬃﬃﬃﬃﬃﬃ the quantities ! ¼ 2 XY , l ¼ X=Y, m ¼ Y=X, where i2 c i2 c i2 c X ¼ 2pþ þ r and Y ¼ 2pþ a þ qa e ae b qb e i2 c 2 . It can be verified that ! is real and takes values rb e from the interval [0, 1]. The ’s are given as

Rs(k)

The ’s entering Eq. (18) are functions of qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ j 2j 1j 1 2 ; j ¼ 0; 1; 2; p pjc ¼ c ¼ pc pc ; 8ðgþ c þ j Þ 1 E 1 1 2 þ ig ¼ þ ; qþ c c 8 gþ gþ gþ c þ 1 c þ 2 c þ 0 1 E 1 1 þ igc þ ; (19) qc ¼ 8 gþ gc þ 2 c þ 1

1 ¼ 11 J1 ðk!Þ;

(a)

0.8

1

ð1 20 Þj 1 2 j : ð 22 1Þ1=2 ð 1 0 Þ2 ð 2 0 Þ2

0.4

1Þ

1=2

0.2

2ð 21

0

J ¼

week ending 19 JULY 2013

PHYSICAL REVIEW LETTERS

0.8

PRL 111, 030403 (2013)

-20

-10

0

k

10

20

30

-30

-20

-10

0

10

20

30

k

FIG. 2 (color online). Characteristic functions R1 and R2 corresponding to the real and the imaginary parts of S12 for ¼ 1. Comparison between the analytical results and the microwave experiment data for the frequency range 10–11 GHz [24–26].

030403-4

PRL 111, 030403 (2013) 3

2.5

PHYSICAL REVIEW LETTERS

(a)

(b)

Analytical Experimental

-1

1.5

P2(x2)

0

0

0.5

0.5

1

1

1.5

P1(x1)

2

2

2.5

Analytical Experimental

-0.5

0

x1

0.5

1

-1

-0.5

0

0.5

1

x2

FIG. 3 (color online). Distributions of the real (P1 ) and the imaginary (P2 ) parts of S12 for ¼ 1. Comparison between analytical results and microwave experiment data for the frequency range 24–25 GHz [24–26].

tilted-stadium billiard [24–26,46] was chosen and the resonator modes were coupled to the exterior via two antennas attached to it. An ensemble of several chaotic systems was obtained by introducing a small scatterer into the microwave billiard and moving it to six different positions [47]. For the determination of the S-matrix elements, a vector network analyzer coupled microwave power into and out of the resonator via the antennas. The frequency range was chosen such that only the vertical component of the electric field strength was excited. Then, the Helmholtz equation is mathematically equivalent to the Schro¨dinger equation of the quantum tilted-stadium billiard. The S-matrix elements were measured in steps of 100 kHz in a range from 1 to 25 GHz and the fluctuation properties of the S-matrix elements were evaluated in frequency windows of 1 GHz in order to guarantee a negligible secular variation of the coupling vectors Wc . More details concerning the experimental setup and the measurements are provided in [24,25]. In Figs. 2 and 3, we test the analytical results with experimental data for the frequency ranges 10–11 and 24–25 GHz, corresponding to a ratio of the average resonance width and average resonance spacing d, =d ¼ 0:234 and =d ¼ 1:21, respectively. The agreement is very good. To conclude, we solved the long-standing problem of deriving the full distribution of the off-diagonal S-matrix elements valid in all regimes. We accomplished this task by introducing a novel route to the sigma model based on the characteristic function. We verified our analytical results with numerical simulations and with experimental data and found excellent agreements, and thus, presented a new confirmation of the random matrix universality conjecture. This work was supported by the Deutsche Forschungsgemeinschaft within the Collaborative Research Centers Project No. SFB/TR12, ‘‘Symmetries and Universality in Mesoscopic Systems’’ and Project No. SFB634 (subproject ‘‘Quantum Chaos and Wave-Dynamical Chaos’’).

week ending 19 JULY 2013

*[email protected] [1] Scattering: Scattering and Inverse Scattering in Pure and Applied Science, edited by R. Pike and P. Sebatier (Academic Press, New York, 2001). [2] R. Blu¨mel and U. Smilansky, Phys. Rev. Lett. 64, 241 (1990). [3] Special issue on Trends in Quantum Chaotic Scattering, edited by Y. V. Fyodorov, T. Kottos, and H.-J. Sto¨ckmann [J. Phys. A 38, 10433 (2005)]. [4] G. E. Mitchell, A. Richter, and H. A. Weidenmu¨ller, Rev. Mod. Phys. 82, 2845 (2010). [5] T. Guhr, A. Mu¨ller-Groeling, and H. A. Weidenmu¨ller, Phys. Rep. 299, 189 (1998). [6] C. Mahaux and H. A. Weidenmu¨ller, Shell Model Approach to Nuclear Reactions (North Holland, Amsterdam, 1969). [7] P. A. Mello, P. Pereyra, and T. H. Seligman, Ann. Phys. (N.Y.) 161, 254 (1985). [8] M. Martı´nez-Mares and P. A. Mello, Phys. Rev. E 72, 026224 (2005). [9] J. J. M. Verbaarschot, H. A. Weidenmu¨ller, and M. R. Zirnbauer, Phys. Rep. 129, 367 (1985). [10] E. D. Davis and D. Boose´, Phys. Lett. B 211, 379 (1988); Z. Phys. A 332, 427 (1989). [11] D. Agassi, H. A. Weidenmu¨ller, and G. Mantzouranis, Phys. Rep. 22, 145 (1975). [12] C. H. Lewenkopf and H. A. Weidenmu¨ller, Ann. Phys. (N.Y.) 212, 53 (1991). [13] M. Baldo, E. G. Lanza, and A. Rapisarda, Chaos 3, 691 (1993). [14] E. Ott and T. Te´l, Chaos 3, 417 (1993). [15] J. A. Folk, S. R. Patel, S. F. Godijn, A. G. Huibers, S. M. Cronenwett, C. M. Marcus, K. Campman, and A. C. Gossard, Phys. Rev. Lett. 76, 1699 (1996). [16] Y. V. Fyodorov and H.-J. Sommers, J. Math. Phys. (N.Y.) 38, 1918 (1997). [17] Y. V. Fyodorov, D. V. Savin, and H.-J. Sommers, J. Phys. A 38, 10731 (2005). [18] I. Rozhkov, Y. V. Fyodorov, and R. L. Weaver, Phys. Rev. E 68, 016204 (2003); 69, 036206 (2004). [19] U. Kuhl, M. Martı´nez-Mares, R. A. Me´ndez-Sa´nchez, and H.-J. Sto¨ckmann, Phys. Rev. Lett. 94, 144101 (2005). [20] U. Kuhl, H.-J. Sto¨ckmann, and R. Weaver, J. Phys. A 38, 10433 (2005). [21] S. Hemmady, X. Zheng, T. M. Antonsen, E. Ott, and S. M. Anlage, Phys. Rev. E 71, 056215 (2005). [22] O. Hul, O. Tymoshchuk, S. Bauch, P. M. Koch, and L. Sirko, J. Phys. A 38, 10489 (2005). [23] M. Ławniczak, O. Hul, S. Bauch, P. Seˇba, and L. Sirko, Phys. Rev. E 77, 056210 (2008). [24] B. Dietz, T. Friedrich, H. L. Harney, M. Miski-Oglu, A. Richter, F. Scha¨fer, J. Verbaarschot, and H. A. Weidenmu¨ller, Phys. Rev. Lett. 103, 064101 (2009). [25] B. Dietz, T. Friedrich, H. L. Harney, M. Miski-Oglu, A. Richter, F. Scha¨fer, and H. A. Weidenmu¨ller, Phys. Rev. E 81, 036205 (2010). [26] B. Dietz, H. L. Harney, A. Richter, F. Scha¨fer, and H. A. Weidenmu¨ller, Phys. Lett. B 685, 263 (2010). [27] C. Ellegaard, T. Guhr, K. Lindemann, J. Nyga˚rd, and M. Oxborrow, Phys. Rev. Lett. 77, 4918 (1996). [28] M. Avlund, C. Ellegaard, M. Oxborrow, T. Guhr, and N. Søndergaard, Phys. Rev. Lett. 104, 164101 (2010).

030403-5

PRL 111, 030403 (2013)

PHYSICAL REVIEW LETTERS

[29] J.-H. Yeh, T. M. Antonsen, E. Ott, and S. M. Anlage, Phys. Rev. E 85, 015202 (2012). [30] J.-H. Yeh, E. Ott, T. M. Antonsen, and S. M. Anlage, Acta Phys. Pol. A 120, A-85 (2011), http://przyrbwn.icm .edu.pl/APP/ABSTR/120/a120-6a-51.html. [31] V. N. Bringi and V. Chandrasekar, Polarimetric Doppler Weather Radar: Principle and Applications (Cambridge University Press, Cambridge, England, 2001). [32] C. Engelbrecht and H. A. Weidenmu¨ller, Phys. Rev. C 8, 859 (1973). [33] G. Casati, F. Valz-Gris, and I. Guarneri, Lett. Nuovo Cimento 28, 279 (1980); M. V. Berry, Ann. Phys. (N.Y.) 131, 163 (1981); O. Bohigas, M. J. Giannoni, and C. Schmit, Phys. Rev. Lett. 52, 1 (1984). [34] T. A. Brody, J. Flores, J. B. French, P. A. Mello, A. Pandey, and S. S. M. Wong, Rev. Mod. Phys. 53, 385 (1981). [35] M. L. Mehta, Random Matrices (Academic Press, New York, 2004). [36] U. Kuhl, R. Ho¨hmann, J. Main, and H.-J. Sto¨ckmann, Phys. Rev. Lett. 100, 254101 (2008). [37] K. B. Efetov, Adv. Phys. 32, 53 (1983); Supersymmetry in Disorder and Chaos (Cambridge University Press, Cambridge, England, 1997).

week ending 19 JULY 2013

[38] J. W. Tepel, Z. Phys. A 273, 59 (1975). [39] J. Richert, M. H. Simbel, and H. A. Weidenmu¨ller, Z. Phys. A 273, 195 (1975). [40] H.-J. Sto¨ckmann, J. Phys. A 35, 5165 (2002); T. Guhr and H.-J. Sto¨ckmann, J. Phys. A 37, 2175 (2004). [41] T. Guhr, in The Oxford Handbook of Random Matrix Theory (Oxford University Press, Oxford, 2010). [42] J. J. M. Verbaarschot, Nucl. Phys. B300, 263 (1988). [43] A. Nock, S. Kumar, H.-J. Sommers, and T. Guhr (to be published). [44] See Supplemental Material at http://link.aps.org/ supplemental/10.1103/PhysRevLett.111.030403 for the explicit expressions of the quantities ij entering Eq. (18) via 1 , 2 , 3 , 4 , and the MATHEMATICA codes for evaluation of the characteristic functions and distributions. [45] Wolfram Research Inc., MATHEMATICA Version 8.0 (2011). [46] H. Primack and U. Smilansky, J. Phys. A 27, 4439 (1994). [47] R. A. Me´ndez-Sa´nchez, U. Kuhl, M. Barth, C. H. Lewenkopf, and H.-J. Sto¨ckmann, Phys. Rev. Lett. 91, 174102 (2003).

030403-6